CHAPTER 1 0

Scalar Field Problems

10.1 INTRODUCTION
In previous chapters, the unknowns in the problem represented components of a vector
field. In a two-dimensional plate, for example, the unknown quantity is the vector field
u(x, y), where u is a (2 xl) displacement vector. On the other hand, quantities such as
temperature, pressure, and stream potentials are scalar in nature. In two-dimensional
steady-state heat conduction, for example, the temperature field T( x, y) is the unknown
to be determined.
In this chapter, the finite element method for solving such problems is discussed.
In Section 10.2, one-dimensional and two-dimensional steady-state heat conduction are
considered, as well as temperature distribution in fins. Section 10.3 deals with torsion of
solid shafts. Scalar field problems related to fluid flow, seepage, electric/magnetic fields,
and flow in ducts are defined in Section lOA.
The striking feature of scalar field problems is that they are to be found in almost
all branches of engineering and physics. Most of them can be viewed as special forms of
the general Helmholtz equation, given by

"-(k:4»
ax ax
+ "-(k, '4» + "-(k '4»
oy' ay az az Z
+ A4> + Q = 0 (\0.1)

together with boundary conditions on ljJ and its derivatives. In the Eq.l0.l,ljJ = ljJ(x, y, z)
is the field variable that is to be determined. Table 10.1 lists some of the engineering
problems described by Eq. 10.1. For example, if we set ljJ = T, k, = ky = k, and A -= 0
and consider only x and y, we get a2T jax 2 + o2T ;oy2 + Q = 0, which describes the heat-
conduction problem for temperature T, where k is the thermal conductivity and Q is
the heat source/sink. Mathematically, we can develop the finite element method for var-
ious field problems in a general manner by considering Eq.10.1. The solution to specif-
ic problems can then be obtained by suitable definition of variables. We discuss here the
heat-transfer and torsion problems in some detail. These are important in themselves,
because they provide us an opportunity to understand the physical problem and how to
handle different boundary conditions needed for modeling. Once the steps are under·
stood, extension to other areas in engineering should present no difficulty. While in other
chapters. both energy and Galerkin approaches were used to derive element matrice, by
Galerkin's approach is used here owing to its greater generality for field problems.

. 306

II...J
.
.
,
:
,
r . ~

TABLE 10.1 Examplt:s of Scalar Field Problems in Engineering

Problem Equation Field variable Parameter Boundary conditions

Heat conduction jPT

k(-+-
iPT) +Q=O aT
Temperature, T Thermal conductivity, k T To,-k-
ilXl ilyl = = %
'n
aT
-k- = h(r - T",,)
'n
Torsion ale+iPO)
(-
iJx!
- +2=0
ill Stress function, (j 9=0

Potential flow
(~J + ~y~-) = 0 Stream function. '" VI = ",,0

Seepage and groundwater flow ,'<I>

k ( - , +--
"<I»
+Q=O
ilx· ily2
Hydraulic potential. 4J Hydraulic conductivity, k <b = -AI
'<I>= 0
-
'n
<b=y
Electric potential .(",n + ''':)
ilx ilr
l
" _p Electric potential. u Permittivity, ~ U U o, - 'nan 0

Fluid flow in duds

(-"w
ilX
",w)
+- - +1=0
I
a}"1
Nondimensional
vc!odty. W
W = 0

Acoustics ii P ill
(- +iJyl
p)
- +k1 p=O Pressure p Wave number,
ilXl p = Po,
(complex) k~ = &i;c 1
I 'P
---= "0
ikpc iln

~
 -------------------------------------------------------------------------------------.

308 Chapter 10 Scalar Field Problems

10.2 STEADY-STATE HEAT TRANSFER

We now discuss the finite element formulation for the solution of steady-state heat-
transfer problems. Heat transfer occurs when there is a temperature difference within
a body or between a body and its surrounding medium. Heat is transferred in the fonn
of conduction, convection, and thermal radiation. Only conduction and convection modes
are treated here.
The heat flow through the wall of a heated room on a winter day is an example of
conduction. The conduction process is quantified by Fourier's law. In a thermally isotropic
medium, Fourier's law for two-dimensional heat flow is given by

q ~
aT
-k- (10.2)
, ay

where T = T( x, y) is a temperature field in the medium, qx and q. are the components

of the heat flux (W1m2), k is the thermal conductivity (W1m. oC),'and aT/ax, aT/ay are
the temperature gradients along x and y, respectively. The resultant heat flux q = qxi
+ q.J is at right angles to an isotherm or a line of constant temperature (Fig. 10.1). Note
that 1 W = 1 J/s = 1 N· m/s. The minus sign in Eq. 10.2 reflects the fact that heat is
transferred in the direction of decreasing temperature. Thermal conductivity k is a

-- material property.
In convection heat transfer, there is transfer of energy between a fluid and a solid
surface as a result of a temperature difference. There can be free or natural convection,
such as the circulation pattern set up while boiling water in a kettle due to hot water ris-
ing and cooler water moving down, or there can be forced convection, such as when the
fluid flow is caused by a fan. The governing equation is of the form
q ~ h(T, - Toc) (103)
where q is the convective heat flux (W/m2), h is the convection heat-transfer coefficient
or film coefficient (W/m 2 • QC), and T, and Too are the surface and fluid temperatures,
respectively. The film coefficient h is a property of the flow and depends on various fac-
tor~ such as whether convection is natural or forced, whether the flow is laminar or tur-
bulent, the type of fluid, and the geometry of the body.
In addition to conduction and convection, heat transfer can also oIXur in the form
of thermal radiation. The radiation heat flux is proportional to the fourth power of the

Isotherm
FIGURE 10.1 Heat flux in two dimensions.
 Section 10.2 Steady-State Heat Transfer 309

absolute temperature, which causes the problem to be nonlinear. This mode of heat
transfer is not considered here.

One-Dimensional Heat Conduction

We now turn our attention to the steady-state heat-conduction problem in one dimen-
sion. Our objective is to determine the temperature distribution. In one-dimensional
steady-state problems, a temperature gradient exists along only one coordinate axis, and
the temperature at each point is independent of time. Many engineering systems fall
into this category.

Governing equation Consider heat conduction in a plane wall with uniform

heat generation (Fig. 10.2). Let A be the area normal to the direction of heat flow and
let Q (Wjm 3) be the internal heat generated per unit volume. A common example of heat
generation is the heat produced in a wire carrying a current] and having a resistance R
through a volume V, which results in Q = ]2 RjV. A control volume is shown in Fig. 10.2.
Since the heat rate (heat flux X area) that is entering the control volume plus the heat
rate generated equals the heat rate leaving the control volume, we have

qA + QA dx = ( q + ~: dx )A (10.4)

Canceling qA from both sides yields

dq
Q~~ (10.5)
dx
Substituting Fourier's law
dT
q~ ~k­ (10.6)
dx

l<fi Right

WaH
face

'\ v face

Heatflowq
-
qA
~A £t:
-
z~q+ dq )
dx dxA
-x

uft RIght
face face
• ·1· d'1
FIGURE 10.2 One-dimensional heat conduction.
~-----------------.

312 Chapter 10 Scalar Field Problems

Galerkin's approach for heat conduction The element matrices will now be
derived using Galerkin's approach. The problem is

:x(k~)+Q~O
TI"o ~ To ql", ~ h(T,. - Too) (10.15)

If an approximate solution T is desired, Galerkin's approach is to solve

(10.16)

for every ~ constructed from the same basis functions as those of T, with ¢J(O) = o.
<p can be thought of as a virtual temperature change that is consistent with boundary con-
ditions. Thus, cP = 0, where T is specified. Integrating the first term by parts, we have

dTi" -
</>k-
dx o
l' 0
kd</>
- -dT
dxdx
dx + l'
0
</>Qdx ~ 0 (10.17)

Now,

</>k~i: ~ </>(L)k(L)~~(L) - </>(O)k(O)~~(O) (10.18a)

Since </>(0) ~ Oandq ~ -k(L)(dT(L)jdx) ~ h(T, - Too). we get

</>k dTi'
dx 0 ~ -</>(L)h(T, - Too) (10.18b)

Thus,Eq.1O.17 becomes

-</>(L)h(T,. - Too) -
l'
o
kd</>
- -dT
dxdx
dx + l'
0
</>Qdx ~ 0 (10.19)

We now use the isoparametric relations T = Nr, etc., defined in Eqs. 10.11-10.14. Fur-
ther, a global virtual-temperature vector is denoted as 'Ii' = ['1'1' '1'2' ... , 'I' I.lr. and the
test function within each element is interpolated as
</> ~ N'i' (10.20)

Analogous to dTjdx = Brr in Eq.l0.13b, we have

(10.21)

Thus, Eq.1O.19 becomes

 Section 10.2 Steady-State Heat Transfer 313

which should be satisfied for all "¥ with "¥ I = O. The global matrices KT and Rare
assembled from element matrices kT and rQ, as given in

k ~k,[ 1 -11] (10.24)

T f,- 1

rQ
_Q,t,{I}
-~-
2 1
(10.25)

When each "¥ is chosen in tum as [0, 1,0, ... ,oy, [0,0,1,0, ... , oy, ... , [0,0, ... ,0,1]T
and since T, = To, then Eq. 10.23 yields

~:: ~) [~:
][ ) [~::~) (10.26)
(KLL + h) iL = ~RL + hT=) - KL:1To
We observe that Eq.1O.26 can be solved for T2 , T3 , • •• , TL. We thus note that the Galerkin
approach naturally leads to the elimination approach for handling nonzero specified
temperature T = To at node 1. However, it is also possible to develop Galerkin's method
with a penalty approach to handle Tl = To. In this case, the equations are as given by
(K" + C) K"
K21 Kn K"
K2L ][ T,T2 ) [(R'
R2 + C7;,) )
(10.27)
[
KLl Ku ... ;K + h) ;,. ~ iR, + hT~)
H

Example 10.1
A composite wall consists of three materials, as shown in Fig. EIO.1a. The outer tempera-
ture is To = 20°C. Convection heat transfer takes place on the inner surface of the waJl with
Tx = 8(XjQC and h = 25 W1m! . dc. Detennine the temperature distribution in the wall.

kl = 2OWlm"e
k, = 30Wfm¢ e
k; = 50Wfmoe
25Wfm 2a C
, h" =
T~ = &XJoe

·IO.l5~lo.15;1
(,)

ttt T!-ll~-;(i)~l:--!;l-0"'2~;"""'0~;4
h, Toc
=20°C

(h)

FIGURE E10.1
---------------------------------------.
314 Chapter 10 Scalar Field Problems

Solution A three-element finite element model of the wall is shown in Fig. ElO.lb. The
element conductivity matrices are

(11=20[ 1 -1] (2) _ ~[ 1

kr 0.3 -1 1 kr - 0.15 -1

(3)_~[
kr - 0.15 -1
1 -1] 1

The global K = kkr is obtained from these matrices as

-1 1 -14 -30 OJ
0
K = 66.7
[o0 -3
0 -5
8-5
5

Now,since convection occurs at node 1, the constant h = 25 is added to the (1,1) location
of K. This results in

K=66.7
1.3-175 -14 -3° OJ0
[ oo -30 -58-55
Since no heat generation Q occurs in this problem, the heat rate vector R consists only of
hToc in the first row. That is.

R = [25 X 800, 0, 0, Op

The specified temperature boundary condition T4 = 20 o e, will now be handled by the

penalty approach. We choose C based on

C = max IKil1 x 104

= 66.7 x 8 X 104
Now, C gets added to (4,4) location of K, while CI4 is added to the fourth row of R. 'The
resulting equations are

°
-3
8 -5
0l{r,} { 25 x800}
0 T2
T,
_
-
0
0
-5 80005 T4 10672 x 104

The solution is

T = :304.6. 119.0, 57.1. 20.0] rO C

Comment. The boundary condition T4 = 20u e can also be handled bv the elimination
approach. The fourth row and column of K is deleted. and R is modi'fied according to
Eq. 3.70. The resulting equations are
·""

Section 10.2 Steady-State Heat Transfer 315

•
66.7
1.375
-1 -14 -3O][T']
T - [25 x
0 gOO]
[
o -3 8 T~ - 0 + 6670
which yields

•
Heat flux boundary condition Certain physical situations are modeled using
the boundary condition
atx = 0 (10.28)
where qo is a specified heat flux on the boundary. If q = 0, then the surface is perfectly
insulated. A nonzero value of qo occurs, for example, due to an electrical heater or pad
where one face is in contact with the wall and the other face is insulated. It is important
to note that the input heat flux qo has a sign convention associated with it: q() is input as
a positive value if heat is flowing out of the body and as a negative value if heat is flow-
ing into the body. The boundary condition in Eq. 10.28 is handled by adding ( -qo) to the
heat rate vector. The resulting equations are

(10.29)

The sign convention for specified heat flux given in Eq. 10.29 is clear if we consider the
heat transfer occurring at a boundary. Let n be the outward normal (in 1-D problems,
n = +x or - x). The heat flow in the body towards the +n direction is q = -k iJT/an,
where iJT/an < O. Thus, q is> 0 and since this heat flows out of the body. we have the
boundary condition q = qo with the stated sign convention.

Comment on forced and natural boundary conditions In this problem,

boundary conditions of the type T = To, which is on the field variable itself. are called
forced boundary conditions. On the other hand. the boundary condition qlx=o = qo, or
equivalently, -k dT/dxx~o = qo is called a natural boundary condition involving the de-
rivative of the field variable. Further, it is evident from Eq.1O.29 that the homogeneous
natural boundary condition q = qo = 0 does not require any modifications in the ele-
ment matrices. These are automatically satisfied at the boundary, in an average sense.

Example 10,2
1
Heat is generated in a large plate (k = 0.8 W/m' cC) at the rale of 4000 W/m· • The plate
is 25 cm thick. The out~ide surfaces of the plate are exposed to ambient air at JOT with a
convective heat-transfer coefficient of 20 W/m" °C Detcnnine the temperature distribu·
tion in the wall.
Solution The prohlem is symmetric about the centerline of the plate. A two·element finite
clement model is shown in Fig. EI0.2. The left end is in~ulated (q = 0) because no heat can
flow across a line of symmetry. Noting that k/f = O.R '.0625 = 12.8. we have
~~---------------------------------------------------.
316 Chapter 10 Scalar Field Problems

I,
I Q = 4000Wlm3

q"'O---il ; ; ttt
h. T~
~6.25Cm+6.25Cm-l
k'" O.8W/m"C
h = 20W/m 2°C
T" = 30"C

fIGURE El0.2

12.8 ~ 12.8
K= -12.8 25.6 o
-12.8 ]
[
o -12.8 (12.8 + 20)
The heat rate vector is assembled from the heat source (Eq. 10.25) as well as due to
convection as

R = [125 250 (125 + 20 X 30W

Solution of KT = R yields

In concluding I-D heat conduction, we note that all element matrices described earlier were
derived using Galerkin's approach. It is also possible to derive these matrices using an
energy approach based on minimizing the functional

nT =
[" (dT)' dx - 10[" QTdx + ~h(TL - Toof
)0 ~k dx (!OJO) •

OneMD;mens;onal Heat Transfer in Thin Fins

A fin is an extended surface that is added onto a structure to increase the rate of heat
removal. A familiar example is in the motorcycle where fins extend from the cylinder
head to quickly dissipate heat through convection. We present here the finite element
method for analyzing heat transfer in thin rectangular fins (Fig. 10.S). This problem dif-
fers from the conduction problem discussed previously in that both conduction and con-
vection occur within the body.
Consider a thin rectangular fin as shown in Fig. 10.6. The problem can be treated
as one dimensional, because the temperature gradients along the width and acrosS !he
thickness are negligible. The governing equation may be derived from the conductlOD
equation with heat source, given by

d( dT) + Q = 0
-
dx
k-
dx
 ,

Section 10.2 Steady-State Heat Transfer 311

Heat dissipation

c::::::::>
Hot gases

FIGURE 10.5 An array of thin rectangular fms.

Convection
heat loss

FIGURE 10.6 Heal f10\\ III a thin rectangular fin.

...-----------------------------------------------------.
~; .

318 Chapter 10 Scalar Field Problems

The convection heat loss in the fin can be considered as a negative heat source
(P dx)h(T - T~)

Ph
~ --(T - T~) (10.31)
A,
where P = perimeter of fin and Ac = area of cross section. Thus, the governing
equation is

-
!«~) ~>T - T~) ~ 0 (10.32)

We present our analysis for the case when the base of the fin is held at To and the tip of
the fin is insulated (heat going out of the tip is negligible). The boundary conditions are
then given by
atx = 0 (10.33.)
atx = L (ID.33b)

The finite element method: Galerkin approach The element matrices and
heat-rate vectors for solving Eg. 10.32 with the boundary conditions in Eqs.I0.33 will
now be developed. GaJerkin's approach is attractive since we do not have to set up the
functional that is to be minimized. Element matrices can be derived directly from the dif-
ferential equation. Let cp(x) be any function satisfying cp(O) = 0 using same basis as T.
We require that

ior' ¢[~(k
dx
dT)
dx
_ Ph (T - Too)] dx
Ac
~0 (10.34)

Integrating the first tenn by parts, we have

¢kdTI' _ r' kd¢dT dx _ Ph r''''Tdx + PhT~ r''''dX ~0 (10.35)

dx 0 io dx dx Ac io Ac io
Using cp(O) "" 0, k(L)[ dT(L)/dx ] = 0, and the isoparametric relations
C, dT dq,
dx~-dg T~NT' "'~N" -~B,T' -~BT"
2 dx dx
we get

(10.36)

We define

h, ~ _ -'
Ph<
Ac 2
l'
-1
NTNdg ~ Ph C, 2 I]
Ac 6 1
[ 2
(10.37.)

L..I
 Section 10.2 Steady-State Heat Transfer 319

or, since P/ A, ~ 2/1 (fIg, 10.6),

h ~ he,[2
3, 1 2
T
I] (1O.37b)

and
(10.380)

or
r
00
~ hToof,
1
{I}1 (10.38b)

Equation 10.36 reduces to

(10.39)
, ,
or
-"I'T(KT + H T) + "I'TR(Xl = 0
which should hold for all 'IJr satisfying '1', = O.
DenotingK;j = (KT + HT)ij,weobtain

Kn K" .:: K~L

KH]lT')
~32 ~33'
[Ku Ku Ku TL
.'
~3
. '
= 1 ) lKHTo)
Roo _ K1:,To
.
(10.40)
,, ,

KuTo
which can be solved for T. These equations incorporate the elimination approach for han-
dling the boundary condition T = To. Other types of boundary conditions as discussed •! I
for heat conduction can also be considered for fin problems. ,, '
'

Example 10.3
A metallic fin, with thermal conductivity k = 360 W1m' °C, 0.1 em thick, and 10 em long,
extends from a plane wan whose temperature is 235"C Determine the temperature distri-
bution and amount of heat transferred from the fin to the air at 20"C with h = 9 W1m2. "c.
I
Take the width of fin to be 1 m.
Solution Assume that the tip of the fin is insulated. Using a three-element finite ele-
ment model (Fig. ElO.3) and assembling K T , H r , R", as given previously, we find that
Eq. 10040 yields

2 -1
360 -1 2
3.33 x 10 2[ 0-1
[
320 Chapter 10 Scalar Field Problems

!i---------2!.---------3~·--------;4Ir_q=O
Tl = 235rl:~C______~3t x3 = lOcm--------~.1
I=O.lcm
w=lm
k=360W/moC

FIGURE El0.3

The solution is
[T" T" T,] ~ [209.8. 195.2. 190.5]'C
The total heat loss in the fin can now be computed as

The loss H, in each element is

H, = h(T.. - T-,o)A,
where A, = 2 X (1 x 0.0333) m2, and T av is the average temperature within the element.
We obtain

HI.", = 334.3 W1m •

Two-Dimensional Steady-State Heat Conduction
Our objective here is to determine the temperature distribution T(x, y) in a long, pris-
matic solid in which two-dimensional conduction effects are important. An example is
a chimney of rectangular cross section, as shown in Fig. 10.7. Once the temperature dis-
tribution is known, the heat flux can be determined from Fourier's law.

I y
I I
I
I
I
I
I
I
1
I I
I I
I I
I I
I I
I I
I I Section a~a

,
I
I :,
FIGURE 10.7 Two·dimensional model for heat conduction in a chimney.
 Section 10.2 Steady-State Heat Transfer 321

q,
dy oQ

1
1-""

FIGURE 10.8 A differential control volume for heal transfer.

Differential equation Consider a differential control volume in the body, as

shown in Fig. 10.8. The control volume has a constant thickness T in the z direction. The
heat generation Q is denoted by Q (W1m3 ). Since the heat rate (= heat flux x area) en-
tering the control volume plus the heat rate generated equals the heat rate coming out,
we have (Fig. 10.8)
q
d q, )
qxdYT+qydxT+Qdxdy'T= ( qx+-dx dYT+ ( q,,+--dyd ,)
dXT
ax - ay
(10.41)

or, upon canceling terms,

aq. aqy (10A2)
-·+-·-Q~O
ax ay
Substituting for qx = -k aTjax and qy = -k aTjay into Eq. 10.42, we get the beat-
diffusion equation

~(kdT) + ~(kdT) + Q = 0 (10A3)

ax ax ay ay
We note that this partial differential equation is a special case of the Helmholtz equa-
tion given in Eq.lO.l.

Boundary conditions The governing equation, Eq. 10.43. has to be solved

together with certain boundary conditions. These boundary conditions are of three types.
as shown in Fig. 10.9: (1) specified temperature T = To on ST, (2) specified heat flux
qn = qo on Sq, and (3) convection q" = h(T - L.) on S,. The interior of the body is de-
noted by A,and the boundary is denoted as S = (Sr + Sq + S,). FUfther,q" is the heat
flux normal to the boundary. The sign convention adopted here for specifying q(l is that
qo > 0 if heat is flowing out of the body, while qo < 0 if heat is flowing into the body.
322 Chapter 10 Scalar Field Problems

.;:::
____--r-'" ,,'
S<:qn=h(T-T,J

A ttt

FIGURE 10.9 Boundary conditions for 2-D heat conduction.

The triangular element The triangular element (Ftg.10.10) will be used to solve
the heat -conduction problem. Extension to quadrilateral or other isoparameteric ele-
ments follows in a similar manner as discussed earlier for stress analysis.
Consider a constant length of the body perpendicular to the x, y plane. The tem-
perature field within an element is given by
T = NITj + N2T2 + N3T.~
or (10.44)
T = NT'
whereN = [t", TJ, 1- g -1J]aretheelemenHhapefunctionsandr = [T j , Tl , T,(
Referring to Chapter 5, we also have
x = Njxj + N2x2 + N3XJ
Y = NjYl + N2 yz + N1J':l (10,45)

• T(x. y)

{=O~l~
Y e= 1

Lx
FIGURE 10.10 The linear triangular element for scalar field problems.
. .,.

Further, the chain rule of differentiation yields

Section 10.2 Steady-State Heat Transfer
•

323
•
aT
~~~~+~-
aTax aTay
a~ ax a~ ay a~
aT aTax aTay
~~~-+~- (10.46)
a'rJ ax d1j ay d1j
or

(10.47)

In Eg. 10.47, J is the Jacobian matrix given by

J ~ [xn Yn] (10.48)

X23 Y23
where Xi} = Xi - X" Yij = Yi - Yi' and IdetJl = 2An where A" is the area of the trian-
gle. Equation 10.47 yields

m)~ r'! ~~ ) (10.49.)

-Yn][10 01 -I]
Xl3 -I
T' (10.49b)

which can be written as

~~) ~ Br T'
where
! aT
ay
(10.50)

(10.51.)

(1O.5Ib)

Galerkin approach~ Consider the heat-conduction problem

-a
ax
(aT)
k~
ax
+ -a ( aT) + Q
By oy
k~ ~ 0 (10.52)

'The functional approach would be based on mininuzl[1g

1r
T
""~
2J4
fJ[k('T)' +k(oT)'
(IX ay
-2QT]dA+ i,.rqoTdS+ is,f_21h(T-r,YdS
• 324 Chapter 10 Scalar Field Problems

with the boundary conditions

T=To on Sr (10.53)
In Galerkin's approach, we seek an approximate solution T such that

(10.54)

for every ~(x, y) constructed from the same basis functions as those used for T and sat-
isfying cP = 0 on ST_ Noting that

q,~(k'T) ~ ~(q,k'T)
oX ax oX oX
_k,q,'T
oX ax
we find that Eq.l0.54 gives

11{[ ~(q,k'T) + ~(q,k'T)]_ [k,q,aT + k,q,'T]}dA

A oX ax ay oy ax ax ay iJy

+ 11 q,QdA ~O (10.55)

- From the notation q" = -k(aTjax) and q, = -k(aTjay), and the divergence theorem,
the first term in Eq.l0.55 above is

11 [:x (q,q,) + :y (q,q,)] dA ~ - /, q,lq,n, + q,n,] dS

(10.56)

where nx and ny are the direction cosines of the unit normal n to the boundary and
qn = q,n x + qyny = q' n is the normal heat flow along the unit outward normal, which
is specified by boundary conditions. Since S = ST + Sq + S,., cP = 0 on S]" q" = qo on
Sq, and qn = h(T - Tx) on Sc> Eq.1O.55 reduces to

- is"r <l>q,dS- is,r .ph(T-T.~)dS-l1(k,q,aT

ax ax
+k,q,'T)dA
ay ay
A

+1.1 q,Q dA ~ 0 (10.57)

Now, we introduce the isoparametric relations for the triangular element such as
T = NT", given in Eqs.l0.47-10.55. Further, we denote the global virtual-temperature
vector as 'It whose dimension equals number of nodes in the fittite element model. The
virtual temperature distribution within each element is interpolated as
q, ~ N", (10.580)
Moreover, just as [aTjax arjay]T = Bl'T~, we have

[a<lax> ' 'ay<I>]T -_ BT"-' (1O.58b)

''T' 'ir' ,
I
Section 10.2 Steady-State Heat Transfer 325

71=0

1)=1

1~ 2

~ = 1 ~=O

FIGURE 10.11 Specified heat flux boundary conduction on edge 2-3 of a triangular element.

Now, consider the first term in Eq. 10.57:

r
JS q
q,q"dS ~ 2: ",Tq"NT dS
e
(10.59)

If edge 2-3 is on the boundary (Fig. 10.11), we have N = [0, 1'/, 1 - 1'/J dS = C2 _3 d1'/,
and it follows that I

( ,.pqo dS = ~ l\ITQOC2_311 NT dT) (10.60.)

JS q e 0

~ 2:, ",T" (1O.6Ob)

where
1 1JT (10.61)

Next, consider
(10.62.)

If edge 2-3 is the convection edge of the element, then

1 </>h(T - Tx) dS ~ ~ "'T[ he,., l' NTN d"]T' -

~ 2: ",ThTT' - 2: ",T" (1O.62b)

Substituting for N = [0, 1'/, 1 - 1]J, we get

hT ~ hf'.T[~ ~ ~]
6 0 1 2
(10.63)

1 1 J' (10.64)
 326 Chapter 10 Scalar Field Problems

Next,

rjk(Oq,oT + oq,oT)dA ~ rjk[oq,O"']!~~)dA (10.650)

JA ax ax ayay iA ax ay aT
oy

~ ~ "'T[ k, 1 BiBTdA ]r (10.65b)

~ ",TkTT' (10.65c)

where
T
kr = krAeBTBr (10.66)
Finally, if Q = Qe is constant within the element,

1j q,QdA ~ ~",TQ, 1NdA ~ ~",TrQ

where

11]' (10.67)

Other distributions of Q within the element are considered in the exercises at the end
of this chapter. Thus, Eq. 10.57 is of the fonn

(10.68)

or
'i"(R oc - R, + RQ) - 'i'T(H, + K,)T ~ 0 (10.69)
which is to hold for a111'" satisfying..p = 0 at nodes on S1 . We thus obtain
KETE = RE (10.70)
where K = LAk r + h t ), R = LAr"<C - rq + rQ)' and superscript E represents ~he
familiar modifications made to K and R to handle T = To on Sr by the eliminatlon
approach. Alternatively, the Penalty approach can also be used to handle T = To.
Example 10.4
A long bar of rectangular cross section, having thermal conductivity of 1.5 Wjm °C is sub-
jected to the boundary conditions shown in Fig. ElO.4a. Two opposite sides are maintained
at a uniform temperature of 180°C; one side is insulated, and the remaining side is subjected
to a convection process with T ~ = 25°C and h = 50 W jm 2 • 0c. Determine the tempera-
ture distribution in the bar.
Solution A five-node, three-element finite element model of the problem is shown in
Fig. ElOAb, where symmetry about the horizontal axis is used. Note that the line of sym-
mctry is shown as insulated, since no heat can flow across it.

~I
-"
Section 10.2

T= 180°C
Steady-State Heat Transfer 327
•
I

T
O.6m k = 1.5W/m"C
!!!
h ~50 Wlm'"C
T" = 25"C

-
~04m~
T = 180"C
(.)

J T= 180°C
4

(I)
(I)
3
~
III 03m

CD h, T~ 1
'X
1 2
q=O
04m---1

(b)

FIGURE El0.4

The element matrices are developed as follows. The element connectivity is defined
as in the following table:

Element 2 3 <-local

1 1 2 3 1
2 5 1 3 global
3 5 4 3 j

We have
B _ _l_[Y2-1
T - det J X32
Y.11
X13
Yl']
X:l
~
I
I[
I !I
,
I
I'
' 328 Chapter 10 Scalar Field Problems
i
For each element,

~.4]
(1) 1 [-0.15 0.15
Dr = 0.06 0 -0.4

~.3J
(2) 1 [-0.15 -0.15
Dr = 0.12 0.4 -0.4

~.41
(3) _ 2[0.15 -0.15
B]" - 0.06 0 -0.4
Then, kr = kAeB~BT yields
ktl) =
,
(1.5)(O.03)BV)'BV)
2 3
~
-~O ]
[ 0.28125 -0.28125
-0.28125 2.28125
-2.0
0
5 , 2.0
3
1'1 _ [ 1.14 -0.86 -0.28125]
kT - -0.86 1.14 -0.28125

- k''I
T
~ [
-0.28125
5
0.28125
-0.28125
-0.28125
4
-0.28125
2.28125
0.5625
3

-~o ]
0 -2.0 2.0

Now the matrices hT for elements with convection edges are developed. Since both ele-
ments 1 and 3 have edges 2-3 (in local node numbers) as convection edges, the formula

h,~h"'[~
0

can be used, resulting in

, 2 3
6 0
2
1 n 5 4 3

h~' ~ [~ ~.25] h\"~[~

0

~.25]
0
2.5 2.5
1.25 2.5 1.25 2.5
The matrixK = ~(kl + h r ) is now assembled. The elimination approach for handling the
boundary conditions T = 180 e at nodes 4 and 5 results in striking out these rows and
0

columns. However, these fourth and fifth rows are used subsequently for modifying the R
vector. The result is

, 2 3
K ~ [ 1.42125 -0.28125 -0.28125]
-0.28125 4.78125 -0.75
-0.28125 -0.75 9.5625
Now the heat-rate vector R is assembled from element convection contributions. The formula

.'f
 Section 10.2 Steady-State Heat Transfer 329

= hTJ2_3 [0 1 lJ
" 2
results in
1 2 3
,~I ~ (50)(27(0.15) [0 1
'1
ond
4 3
(50)(25)(0.15) (
r~) =
2 0 1 1 I
Thus,
1 2 3
R ~ 93.75[0 1 2]'

In the elimination approach, R gets modified according to Eq. 3.70. Solution of KT = R

then yields
[T" T2 , T1J = [124.5, 34.0, 45.4JoC
Note: A large temperature gradient exists along the line connecting nodes 2 and 4.
This is because node 4 is maintained at 180°(:' while node 2 has a temperature close to the
ambient temperature of T~ = 25°C because of the relatively large value of h. This fact im-
plies that our finite element model should capture this large temperature gradient by hav-
ing sufficient number of nodes along line 2-4. In fact, a model with only two nodes (as
opposed to three as used here) will lead to an incorrect solution for the temperatures. Also,
with the three-element model considered here. heat-flow values (see computer output) are
not accurate. A more detailed model is necessary. •

It is also noted that a thermal-stress analysis can now be perfonned once the tem-
perature distribution is known, as discussed in Chapter 5.

Two-Dimensional Fins
In Fig. 10.12a, a thin plate is receiving heat from a pipe and then dissipating it to the
surrounding media (air) by convection. We may assume that the temperature gradients
are negligible in the z direction. Thus, the problem is two-dimensional. Our interest is
to detcrmine the temperature distribution T(x, y) in the plate. The plate is the fin here.
Considering a differential area dA, the convection heat loss from both lateral surfaces
of the fin is 2h(T - T"",) dA. Treating this heat loss as a negative heat source per unit vol-
ume, Q =: -2h(T - TYo)/t. where t =: thickness of the plate. Equation 10.43 yields the
differential equation for two-dimensional fins. namely.

'--(kaT)
ax ax
+ '--(kaT) _C(T - T~) + Q ~ 0
Ely Ely
(10.71)

where C = -2h/t. Another example of a two-dimensional fin may be found in elec-

tronic packaging. The thin plate shown in Fig. lOJ2b is subjected to a heat source from
the surface underneath gcnerated from electronic chips or other circuitry. Pin fins are
attached to the top surface to dissipate the heat. As shown in the figure, the plate may
 ,,
Ii I
~ '-

330 Chapter 10 Scalar Field Problems

Convection

- x •
..... Hot
gas, Tg

T(r,y)
(,j
y
pin fin
h pin hn

- t23 ,--
t23
,
, --,'
-
' x

tzI tzI tzI hplate

Heat source
(bj

FIGURE 10.12 Two-dimensional fins.

be considered as a two-dimensional fin with higher convective heat-transfer coefficients

where the pin fins are attached. In fact, these coefficients may be related to the fin size
and material. Maximum temperature at the surface of the chip will be of importance in
this analysis. The conductivity matrix k in Eq.1O.66 and the right-hand-side heat-rate vec-
tor TQ in Eq.1O.67 get augmented by the matrices

+ ~;[: ~ nand +CToo ~'m (10,72)

respectively.

Preprocessing for Program Heat2D

The input data file for program HEAT2D can be created, in most part, using the MESH-
GEN program. Mesh generation is as usual. Treat specified temperatures as "constrained
degrees of freedom," nodal heat sources as "loads," element heat sources as "element
characteristics" (enter zero if there are none), and thermal conductivity as "material
 Section 10.3 Torsion 331

property." The only thing that remains is the heat flux and convection boundary condi-
tions along the edges; for this, simply edit the data file that you have created and enter
this information as per the format of Example 10.4 that has been provided at the end
of this chapter. Note that in heat conduction each node has only one degree of freedom.

: :
10.3 TORSION
Consider a prismatic rod of arbitrary cross-sectional shape, which is subjected to a twist-
ing moment M as shown in Fig.lO.13. The problem is to determine shearing stresses T.tZ
and T y<: (Fig. 10.14) and the angle of twist per unit length, a. It can be shown that the so-
lution of such problems, with simply connected cross sections, reduces to solving the
two-dimensional equation

inA (10.73)

onS (10.74)
where A is interior and S is the boundary of the cross section. Again, we note that
Eq.l0.73 is a special case of Helmholtz's equations given in Eq. 10.1. In Eq. 10.74, 9 is ,!
called the stress function, since once 9 is known, then shearing stresses are obtained as
oe = -Ga-
ae (10.75)
Txz = Ga~ T.
ax
oy y-

with a determined from

M = 2Ga 1f edA (10.76)

where G is the shear modulus of the material. The finite element method for solving
Eqs.l0.73 and 10.74 will now be given.

r x

FIGURE 10.13 A rod of arbitrary cross sectlon subjected 10 a torque.